Pass HPA autoscaling behavior through the Helm chart

[sdolidze]

Goal

Allow Helm users to configure Kubernetes HPA scale-up and scale-down behavior. Operators can then prevent rapid replica growth from multiplying metadata-database connection demand during pressure.

Context

The chart uses autoscaling/v2 and exposes replica bounds plus CPU and memory targets. It does not render spec.behavior, so users cannot configure stabilization windows or scaling policies through values.

Settings model

Field	Type	Default	Meaning
autoscaling.behavior	Kubernetes HorizontalPodAutoscalerBehavior object	omitted	Native autoscaling/v2 scale-up and scale-down behavior

Chart changes

• Add a documented optional autoscaling.behavior value.
• Render the object under spec.behavior with toYaml.
• Regenerate the chart README from the values documentation.

Verification

1. ct lint --all passes.
2. Default helm template output is unchanged.
3. A values fixture with scale-up and scale-down policies renders valid autoscaling/v2 YAML.
4. Nested policies, selectPolicy, and stabilizationWindowSeconds pass through without chart-specific restrictions.

Files

• charts/lightdash/values.yaml
• charts/lightdash/templates/hpa.yaml
• Chart documentation and render tests

Out of scope

• Choosing default rate limits for operators.
• Changing replica, CPU, or memory defaults.
• Adding application-level backpressure.

[linear-code]

PROD-9787